 It's work called Map Hub, and it's about annotating high-resolution digitalized historical maps. And so we built a specific application for this purpose called Map Hub, and it uses great technology. One of my colleagues developed back in Austria, Reiner Simon, and Konit continues developing under the name Annotorious. So after six minutes and 30 seconds, I will switch over to him and he will tell you something about Annotorious. So why do I think that historical maps are such a great subject for annotations? A while ago, I was at the Library of Congress in one of their exhibitions where they showed the Baltimore map as one of the maps, and it was just a spin where they live in, which was pretty hard for most American people with this map because it's from 1508. But pretty interesting for Europeans or Asians in this case. This was also confirmed in user studies. I did it at Cornell. I noticed that the students we invited for our experiments, they also spent considerable time investigating this map and other maps. We had students transcribing Latin inscriptions. We had students trying to identify the locations of Taiwan and Japan on this map. And we had students who were just amazed by the fact that this map shows an elephant in South Africa and monsters in the Pacific and Atlantic Ocean and they gave us this information, this factual knowledge as annotations. So I think the three major benefits of thinking of annotations in the context of historical maps are first, it's a way to engage people to share the factual knowledge and the stories with others so they can attach this directly to the map. Second, for institutions like the Library of Congress, annotations are a way to collect knowledge and data. So I, please correct me if I'm wrong, but I think the fact that the Baltimore map shows an elephant is not part of the Library of Congress metadata. So if you search for maps showing elephants, you won't find anything, right? But if you get the information as part of annotations, this can become part of the metadata records in an institution and be the basis for search and retrieval. Library of Congress is already doing this with images, they're using Flickr comments and they're integrating social media comments into the metadata records if they think that they're factual and useful for their purpose. So I think you could do exactly the same for historical maps as well. And third, I think annotations are an excellent way to get historical maps out of their silos. Most institutions still have them back in their, whatever, storage mechanisms and connect them with other resources on the web and I will show you how we did this. So we implemented MAPUP, it's an open source project available on GitHub. If you go to mapup.github.com, you will see a seven minute video showing all the features we have at the moment, I don't have the time now, so I will just briefly summarize what we did. Again, MAPUP is an online application for exploring and annotating high-resolution historical maps. At the moment, it's bootstrapped with approximately 6,000 maps from the Library of Congress, but the plan is to extend this. And the core features we supported at the moment are first, georeferencing maps and creating map overlays from historical maps on top of modern mapping systems like Google Maps or Google Earths. The second feature we support is textual annotation plus semantic tagging. I will explain later what this is. Since we support semantic tagging, we can also provide, in my opinion, a very interesting way of multilingual search, so you can search in Klingonian and Estonian Esperanto for maps because people provide this knowledge. I will show you later. And last but not least, all annotations created in MAPUP are some first-class web resources, so you can just say HTTP GET my annotation and you have it in your system. So what does georeferencing mean? Georeferencing means that people identify places on MAP and just tell us what those places are. So they say, in this case, okay, these places are done in Cuba. But underneath the user interface, what we do is we are essentially creating an association between an XY point on a raster image map and the first-class web resource in georenames. So we're connecting a point on a map with a web resource, so we have a kind of a semantic tag. As soon as we have three or more georeferences for a map, we can compute an approximate translation model between raster images and spherical micarta projection, in the case of Google. We overlay historical maps on Google Earth, Google Maps, and this opens a whole new world for cool features. So you could, for instance, draw annotations on Google MAP and view them on historical maps or the other way around. So one of the things I was thinking of, you can, for instance, draw your favorite running route on Google My Maps and just see what this area was in the 18th century we were running through. So there are plenty of opportunities. The second feature we annotated, we've implemented is textual annotation. So if you know something about this map, you can select specific regions on the map. You can use several selectors, polygons, rectangles, whatever, and you can just add your knowledge. And in this case, I created an annotation saying that this area shown on this map is the Strait of Gibraltar, formerly known as the Pillars of Hercules, which is the reason for this inscription. What we are doing while the users are annotating and typing their text is we are on the fly analyzing their text and we are proposing tags, semantic tags. This means the labels you see here, like Mediterranean Sea, Strait of Gibraltar, et cetera, are not strings. Those are links to Wikipedia. And the user can accept or reject those tags. So users essentially create positive and negative tagging relationships to Wikipedia, which we use in this case, while they are writing their notations. So this is a very precise way of getting unambiguous references to clearly defined concepts instead of strings. I don't have much time to talk about this more. I just would like to refer to a study we did at Cornell last year, because we were interested how this semantic tagging behaves compared to traditional label-based tagging, as you know it from Flickr. We had an in-lab experiment with 24 students and our findings was that it doesn't affect the tag production, the types and categories of tags you get, and it's also not more frustrating for users than label-based tagging. So it's a pretty cool finding, and if you have semantic tags in your system, as we do, you can do pretty cool stuff. You can, for instance, if you have a link to Wikipedia, just translate it to a dvbida link, in all the translations you get from dvbida, and then enable multilingual search if you just have it in your index. So if some user told us that this is the Mediterranean Sea, you can then just search for this map using the Japanese Mandarin, whatever language is supported by Wikipedia. Last but not least, we share each annotation using the W3 Open annotation API. Rob talked about it before, so you just say HTTP GET, my cheer reference, HTTP GET, my commentary annotation, and you have it in your system.